DAMAGE LEVEL PREDICTION OF MULTI-STORY STEEL STRUCTURE IN SUMATRA USING BACKPROPAGATION NEURAL NETWORK

Abstract

Sumatra is one of the Indonesia islands that is prone to earthquakes both tectonic and volcanic. The research aims to predict the damage level of a multi-story steel structure due to the earthquake in Sumatra Island using the Backpropagation Neural Network (BPNN). The study used the steel structure building that received earthquake loads from ten capital cities of the province on Sumatra Island. The structure analysis used the finite element software while the BPNN method used the MATLAB Programming. The input data were the responses of the structure such as displacement, velocity, and acceleration while the output was damage level of the steel structure model. The model of BPNN has the potential accuracy to predict the damage level of steel structural more than 95%. According to the simulation result, 98,5% data could be predicted correctly by the BPNN method, and the best Mean Squared Error (MSE) is 0.028. These results have shown that BPNN can predict the damage level of multi-story steel structure in all the capital cities of the province on Sumatra Island.